1. Across Breed EPD and multi-breed genetic evaluation developments
Larry Kuehn
USDA, ARS, U.S. Meat Animal Research Center
The USDA is an equal opportunity employer.

2. Across breed EPD program
Program has been in place since 1993
Birth, weaning, yearling weight; milk
Carcass traits added in 2008
Carcass weight added in 2015
Develop factors that allow producers to compare genetic merit of bulls across breeds
Data from USMARC Germplasm Evaluation Program

3. 2015 ABEPD factors

4. Genetic Trends for Yearling Weight, lb
Adapted from Spring 2015 Genetic Trends from Breed Associations
and 2015 AB-EPD factors

5. Limitations of ABEPD program
Factors do not account for difference in heterosis in commercial cross
Angus vs. Simmental bull with Angus females
Additive factors not adequate to accommodate many traits
Calving ease, heifer pregnancy, etc.
Yearly updates not sufficient for continuously updated national cattle evaluations

6. Possible solutions Web-based decision support Multibreed evaluation
Current focus of a collaboration/grant
KSU, UNL, CSU, USMARC, Bruce Golden
Could combine with bioeconomic simulation
Heterosis, trait scaling, continuous updates all possible
Multibreed evaluation
Current implementation by American Simmental Association/International Genetic Solutions (ASA/IGS)

7. Obstacles to full Multibreed
Merger of multiple breed databases
Structures are often very different
IDs duplicated in several breeds (but not known as duplicates)
Difficult to resolve
Standardized ID system would help
Cooperation between database curators
Breed associations
Genetic prediction ‘centers’
Individual producers/commercial entities

8. Multibreed obstacles Estimating population parameters
Direct and maternal heterosis
Direct and maternal additive breed effects
Field data usually not suitable
Contemporary groups structure
Will discuss further
Confounding between heterosis and breed
Amount of crossbred data relative to purebred (depends on classification of ‘purebred’)
Research data useful here

9. Contemporary groups
In order to estimate breed differences from field data, we need contemporary groups that include purebreds of both breeds
Rarely occurs; often breeds are in different groups
Even when crossbreds and purebreds are in the same group, direct comparisons are not possible without adjusting for heterosis (requires good estimates of heterosis)

10. Estimation of heterosis
In order to estimate heterosis from field data, we need groups with crossbreds and purebreds of both parental breeds
Rarely occurs; usually crossbreds are in a different groups
Even when crossbreds and purebreds are in the same group, typically purebreds of only one of the breeds are present

11. Estimating breed differences
Contemporary group
All mated to mature purebred Limousin cows
What was the cause of the different averages?
Heterosis
Sire breeding value
Breed differences
Sire 1: Limousin
WW Avg: 650 lb
Sire 2: Angus
WW Avg: 675 lb
Sire 3: Lim-Flex
WW Avg: 670 lb
YES
(and no)

12. Estimating breed differences
Problem can be improved with more sires in group still other considerations that are difficult to address
Reciprocal matings
Biased sampling of sires from other breeds
Heterosis still difficult to separate from breed
Were calves really treated the same?

13. Multibreed model
Prior estimates of breed effects and heterosis essentially required
Source of information most likely from research data
GPE program is designed to estimate breed differences from current industry samples

14. Proposal
Use breed differences from GPE to parameterize multibreed model currently in use by American Simmental Association
Provide both breed effects and heterosis
Can basically pre-adjust data for breed composition
Need to consider where the programs are different and how to accommodate differences

15. GPE Target Population Structure
AI Sires:
AN, HH, SM, CH, AR, LM, GV, SH, BN, BM, MA, BR, CI, SG, SA, BV, SD, TA 
PB & F1 Steers
PB & F1 Bulls
PB & F1 Heifers 
Natural Service PB, F1, & F12 Steers & Heifers

16. Across breed differences

17. Potential problems
Currently, priors for ASA/IGS model are for breed by year effects
GPE analysis based on sampling from industry sires and adjusting solutions to the EPDs of the sampled bulls
In essence using breed genetic trends to adjust solutions from GPE data
Only Hereford and Angus bulls sampled throughout GPE
Interpolation likely necessary

18. Potential problems New trait development
Currently summarize whole GPE database for weight traits and carcass traits as part of the ABEPD process
Still missing CED, CEM, stayability and heifer pregnancy that are reported for several breeds
Multinomial distributions of these traits will require some form of scaling from GPE to multibreed

19. Potential problems Heterosis
While heterosis is reported as part of GPE, prior to current continuous sampling protocol, most estimates were based on Angus x Hereford crosses
One goal of current program is to estimate breed-specific heterosis
Important for multibreed
Still far from complete

20. Possible solutions Breed x year solutions
Need to examine what we can do to fill in the years
Current ASA method places a high correlation among yearly estimates
Possibly can reference old solutions and newer solutions and interpolate
Adjust within years by within group EPD differences
Would exclude genetic trend (circular logic)

21. Possible solutions New trait development
Have already prototyped CED with UNL collaboration
For ABEPD system, need to put all breeds on the same scale then transform to breed of interest
Scaled EPD by additive SD of EPD
Similar methodology could be applied to heifer pregnancy and stayability
Records may be limited on some breeds in GPE, but will grow over time

22. Possible solutions Breed specific heterosis
UNL collaboration also developed to begin exploring breed and breed-type specific heterosis (Schiermiester et al., 2015)
Fitting main effect of heterosis (or breed-type) and random breed specific heterosis effects
Resolution not great yet, but evidence of breed-type specific heterosis for weight traits
Working on increasing crosses between 7 largest breeds and of those breeds with all others

23. Future plans
Continue trait development and examination of breed effect model implications
Important training in NCE for graduate student development

24. Conclusion
We think the GPE program and the multibreed model are a natural fit
Will improve the base adjustment among current members of IGS
Can help to transition to mating and selection decision support programs

25. Acknowledgements
Bruce Golden Wade Shafer Warren Snelling Mark Thallman Matt Spangler Lauren Hyde

26. Discussion and questions?
Mention of a trade name, proprietary product, or specific equipment does not constitute a guarantee or warranty by the USDA and does not imply approval to the exclusion of other products that may be suitable.